On Jun 28, 2000, Akim Demaille <[EMAIL PROTECTED]> wrote:
> Why would someone set host != build
> and not expect cross-compiling?
I'm taking about the case in which build is not specified. In this
case, the user may have meant the old behavior, in which --host would
set the default for build and target, and s/he only wanted to change,
say, i686-pc-linux-gnu to i686-redhat-linux-gnu.
> What's wrong with
> (1) cross_compiling = build != host
Nothing. As long as build is specified. Comparing a guessed build
with a given host would probably not do what the user intended. Doing
otherwise would actually break build scripts that set --host just to
override the guess in order to add a vendor name. Those will
eventually have to be changed to set --build, but may prefer to set
both --build and --host to be compatible with both old and new
packages.
> Are users really used to use --host alone?
Vendor build scripts usually do that. Another relevant case is
--host=i586-pc-linux-gnu on a i686-pc-linux-gnu. This needs not be
taken as cross-compilation if it doesn't have to.
> I fear the `maybe'.
Don't. Trust me :-)
Now it's better than what it used to be, in most cases. Only for the
few people that set host but not build, this might become a problem.
Alexandre> There's no way for configure to correctly guess the build
Alexandre> platform, in general, when it's told to use a cross
Alexandre> compiler.
> I don't understand this too well. Are you saying that since people
> are used to CC=sun4-cc, config.guess will actually config.guess sun4
No. It may completely fail to guess (e.g., on
sparc{,v9}-unknown-linux-gnu), guess something slightly different
(e.g., i686-pc-linux on a i686-pc-linux-gnu) or just do the right
thing.
--
Alexandre Oliva Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist *Please* write to mailing lists, not to me
